electrical & comput engineering · 3.1. description slik silicon apds are state-of-the-art...
TRANSCRIPT
The Johns Hopkins University
p. 23
(NASA-CR-190561) PERFORMANCE MEASUREMENTRFSULTS FOR A 220 Mbps QPPM OPTICALCOMMUNICATION RECEIVER WITH AN FG/G SLIK APDInterim Progress Report, 16 Oct. 1991 15Jul. 1992 (JHU) 23 p G3/32
N92-29658
Unclas0109254
ELECTRICAL& COMPUTENGINEERING
https://ntrs.nasa.gov/search.jsp?R=19920020415 2020-07-16T18:28:30+00:00Z
Performance Measurement Results tor a220 Mbps QPPM Optical Communication
Receiver with an EG&G Slik APDFrederic M. Davidson and Xiaoli Sun
Department of Electrical and Computer EngineeringThe Johns Hopkins University, Baltimore, MD 21218
- Progress report on Grant NAG5-510'Reduced electrical bandwidth receivers for direct detection
4-ary PPM optical communication intersatellite links'for the period Oct. 16, 1991 to July 15, 1992.
July 1992
Performance Measurement Results for a220 Mbps QPPM Optical Communication
Receiver with an EG&G Slik APD
Frederic M. DavidsonXiaoli Sun
Department of Electrical and Computer EngineeringThe Johns Hopkins UniversityBaltimore, Maryland 21218
July 1992
1. Introduction
The performance was measured of a 220 Mbps quaternary pulseposition modulation (QPPM) optical communication receiver [1] witha 'Slik' silicon avalanche photodiode (APD) and a widebandtransimpedance preamplifier in a small hybrid circuit module. Thedetails of the QPPM encoder and receiver electronics are describedin Reference [1]. The receiver performance had been poor due to thelack of a wideband and low noise transimpedance preamplifier. Withthe new APD preamplifier module, the receiver achieved a bit errorrate (BER) of 10~6 at an average received input optical signal powerof 4.2 nW, which corresponds to an average of 80 received (incident)signal photons per information bit.
2. Test Setup
The test setup is shown in Figure 1. The BER tester transmitter(Microwave Logic BERT-660Tx) generated pseudorandom binarysequences and served as a binary data source. The QPPM encoderconverted every two bits into a PPM word which in turn modulatedthe laser transmitter. The optical signal was attenuated andphotodetected by the silicon APD. The noise corrupted PPM sequences
PRECEDING PAGE BLANK NOT FILMED
output from APD preamplifier were subsequently processed anddemodulated by the QPPM receiver. The BER tester receiver(Microwave Logic BERT-660Rx) compared the recovered binarysequences with the source binary data and determined the receiverBER.
The laser transmitter consisted of an AIGaAs laser diode(Mitsubishi ML5702A) operated at a wavelength of about 820nm. Thelaser diode driver and temperature control unit was made by NASAGoddard Space Flight Center. The laser diode was biased slightlybelow the threshold current (47mA) and the temperature wasmaintained at 25 °C. The average output optical power at the exit ofthe laser beam collimation lens was about 3.6 mW. Figure 2 showsthe laser pulse shapes measured with a high speed photodiode (Opto-Electronics PD10). The photodiode had an inverting output due to itsinternal circuit configuration. The pulse rise and fall times from10% to 90% pulse height were 330 and 500 ps, respectively. Theaverage pulse width was about 2.1 ns. It was also found thatpulsewidths varied by as much as 8.5% from one PPM word toanother, which might have caused some degradation in receiverperformance.
The propagation losses in free space were simulated by a set ofneutral density attenuators. They were also used to adjust the inputoptical signal power to the receiver. The total attenuation of theattenuators was about 53 dB.
The receiver section of the optics consisted of a interference filterand a focusing lens. The interference filter was used to block outbroad band laser diode noise and background radiation noise. Thebandwidth of the filter at full width at half maximum (FWHM) was9.0 nm and the peak transmission coefficient at the centerwavelength was about 38%. The focusing lens was a 21X microscopeobjective lens with numerical aperture of 0.5.
3. Slik APD and Preamplifier Module
3.1. Description
Slik silicon APDs are state-of-the-art devices recently developed byEG&G Canada [2] which feature a 'super low ionization coefficients,1
keff =0.005 as compared to keff =0.020 of the commercial gradedevices. Typical quantum efficiency at 800 nm wavelength is aboutT|=90%. The diameter of the active area of the APD is 100 jim. Ahybrid circuit module was made by EG&G Canada which integrated aSlik APD with a lownoise and wideband (-1.0 GHz) transimpedancepreamplifier (Anadigics ATA12000 [3]). The transimpedance of thepreamplifier is 1.5 kQ. The rated noise current density is about 5.0pA/VHz at 500 MHz and increases with frequency. The preamplifieralso contains an automatic gain control (AGC) circuit although theAGC threshold current is relatively high (100 u.A) as compared withour normal operation signal level (-0.5 u,A at 5 nW input opticalsignal level). The APD high voltage bias supply consisted of aprogrammable DC to DC converter (Analog Modules 522-2). It had aninternal temperature compensation circuit and the output ripple was5 mV rms.
3.2. Test Results
The electrical characteristics of the APD preamplifier module (EG&GCanada C30964E CD1796, serial number 0011) were measured beforeuse in the receiver. The bandwidth of the APD preamplifier modulewas determined by measuring the noise power spectrum of the APDpreamplifier output while shining a relatively strong white lightonto the APD active area. The APD was biased close to thebreakdown voltage (-351 V). Two amplifiers (MiniCircuit ZFL1000-LN and ZFL1000, total gain 42dB) were used to further amplify thepreamplifier output signal before being input to the spectrumanalyzer. Figure 3 shows the measured spectrum. The 3dB bandwidthwas found to be about 930 MHz.
The preamplifier noise current density was found by measuring thenoise power spectrum while keeping the APD completely in the dark.The only noise sources under this condition were the APD darkcurrent and the preamplifier circuit noise. The APD dark currentnoise was found so small that it caused little change in the noisespectrum when the APD bias voltage varied from well below to nearthe breakdown voltage. Figure 4 shows the measured noise powerspectrum. At 200 MHz, the noise density was -112.8 dBm/Hz.Considering the 42 dB gain of the amplifiers, the noise powerdensity at the preamplifier output was -155.8 dBm/Hz or 2.63X10-19 Watts/Hz. Multiplying it by 50Q and taking the square root, thevoltage noise density was 3.62 nV/VHz. Dividing it by thetransimpedance (1.5 kQ), the current noise density was found to be2.4 pA/VHz. The spectral noise density measured at 800 MHz wasroughly 8 dB greater than that at 200 MHz. Consequently, the currentnoise density at 800 MHz was 6.3 pA/VHz. These measurement dataagreed well with those given in the data sheet [3].
The APD ionization coefficient, keff, was found by measuring the APDoutput excess noise factor under a relatively strong CW incidentlaser light such that the APD excess noise became dominant. TheAPD excess noise factor, F, is defined as the ratio of the meansquare to the square of mean of the APD gain and is related to theAPD ionization coefficient by [4]
F = keffG + (2- ) (1-keff) (1)
where G is the average APD gain. The excess noise factor is also thesolution to the equation [5]
d<i2>ncw 9 Fr23fcwdf = 2qFG hf
where d<i2>ncw/df is the spectral noise current density, q is theelectron charge, PCw is the incident CW laser optical power, and hf isthe photon energy.
The spectral noise current density was measured directly using aspectrum analyzer, similar to how the preamplifier noise currentdensity was measured. The noise power spectral density was alsomeasured with a RF power meter and a known bandwidth lowpassfilter. The results were consistent with those measured with thespectrum analyzer. The incident CW laser optical powers were heldat 50 nW and 100 nW, which were well below the preamplifier AGCthreshold.
The average APD gain was determined by modulating the laser diodewith a QPPM signal and measuring the pulse amplitude, Vp-p, of theoutput from the APD preamplifier. The average APD gain is thesolution to the equation [5]
(3)
where Pp-p is the peak-to-peak incident optical signal power, Rf isthe preamplifier transimpedance, and A is the total gain of theamplifiers used after the preamplifier. The laser ON-OFF extinctionratio was kept high such that Pp.p was equal to the average opticalsignal power, Pav, divided by the PPM signal duty cycle, 25%. Theaverage power was measured directly with an optical power meter.The APD gain was fixed at G =100 during the measurements.
The APD ionization coefficient was obtained by substituting (3) and(2) into (1). The resultant ionization coefficient was keff=0.02. It isnoted that the measured keff was four times greater than that of aSlik APD described in [2]. The discrepancy is yet to be explained.
Figure 5 shows the average pulse shape output from the APDpreamplifier at 9.45 nW average input optical signal power.Pulseshape distortions due to the APD and the preamplifier areobvious when comparing Figure 5 to Figure 2.
4. Receiver Frontend
4.1. Amplifiers following the APD Preamplifier
Figure 6 shows a circuit diagram of the receiver frontend, whichshould supersede Figure 4 in Reference [1]. An Avantek UTO511/505cascade amplifier (5-500 MHz, 28 dB) was chosen as the first stage.The undershoot due to its relatively high lower cutoff frequencycounteracted the residuals of pulse trailing edges at the bottom ofthe trace and consequently helped to reduce intersymbolinterference caused by the residual. The signal was furtheramplified by another linear amplifier (MiniCircuit ZFL1000-LN, 0.1-1000 MHz, 23 dB). A variable attenuator was used between the twoamplifiers to adjust the total gain. Figures 7 and 8 show the outputsignal waveforms. The signal was then split into the timing recoverycircuit and the QPPM detection circuit.
4.2. Modification to the Timing Recovery Circuit
The original timing recovery circuit could tolerate only about 6 dBinput signal dynamic range, which had caused great difficulties inoptimizing the average APD gain and measuring the receiver BER vs.input optical signal power. The GaAs phase lock loop (PLL) chip(GigaBit Logic 16G040) for the PPM slot timing recovery was foundto have too narrow an input signal dynamic range. We improved thisby adding a high speed comparator as a limiter before the PLL chip.The detailed circuit diagram of the modified timing recovery circuitboard is shown in Figure 9, which should supersede Figure 6 inReference [1]. The original Avantek GPD311/321 linear amplifier [1,p. 7] was also replaced with a wider dynamic range AvantekGPD321/330 amplifier. The resultant dynamic range of the modifiedreceiver increased to at least 20 dB as measured by connecting theQPPM encoder output directly to the receiver.
Two synchronization lock indicators were also added, one for theQPPM slot clock PLL and one for the QPPM word clock PLL. The outputof the op-amp of the active loop filter of each PLL was first
buffered and then connected to a LED indicator. When a PLL was outof lock, the op-amp almost always drifted into saturation in onedirection with a constant output voltage which turned on the LEDindicator.
4.3. Raised Cosine Filter
An approximate raised cosine filter was installed in place of theoriginal tapped delay line matched filter. The filter consisted of alinear amplifier (Avantek UTC2033) with a 20 pF shunt capacitor atthe input, as shown in Figure 6. The input impedance of the amplifierand the shunt capacitor formed a lowpass filter which had afrequency response similar to a RC filter with a 3dB bandwidth ofabout 300MHz and a stopband (-20dB) of about 500 MHz. Theattenuators at the input and the output of the linear amplifier werenecessary to reduce the effect of impedance mismatching due to theadded shunt capacitor. The output from the raised cosine filter wasamplified again by a high power amplifier (Trontech P500A-3) todrive the subsequent QPPM detection circuit. Figure 10 shows theaveraged waveform of the power amplifier output. The waveformlooks similar to an ideal raised cosine pulses (inverted).
It should be pointed out that the lowpass filter we constructed wasprobably not a very good approximation to a true raised cosine filteras described in [6]. The receiver performance is expected to improveif one uses a better approach such as a Bessel lowpass filter or atransverse (tapped delay line) lowpass filter with similarbandwidth.
5. Measurement Results
The receiver BER was measured as a function of the average inputoptical signal power and the results are presented in Table 1 andFigure 11. The receiver achieved a BER of 10-6 at an average inputoptical signal power of 4.2 nW (-53.8 dBm), which corresponded to80 incident photons per information bit. The measured optimalaverage APD gain was G=110.
8
The theoretical performance of an ideal QPPM receiver with aperfect rectangular input pulse shape and a matched filter is alsoplotted in Figure 11. The parameter values used are listed in Table 2.The laser ON-OFF extinction was assumed high (300) since the laserwas biased below the threshold current. The APD bulk and surfaceleakage currents were taken to be 1.0 pA and 15 nA, which weretypical values for silicon APDs. The noise spectral density of thepreamplifier was assumed to be constant over the entire bandwidthand equal to 5.1 pA/VHz (1.5 KQ feedback resistance at anequivalent noise temperature of 700°K).
The measured input optical signal power at BER=10-6was about 1.5dB higher than that of an ideal QPPM receiver. Considering the losses(-0.5 dB) due to the use of a raised cosine filter and nonrectangularinput pulse shape, the difference between the theoreticalcalculation and the measurement should have been about 1.0 dB. Webelieved that this difference is mainly caused by the intersymbolinterference from the ringing and long trailing edges of the pulsesoutput from the APD preamplifier. The calculated optimal averageAPD gain at BER=10'6was about 36% higher than the measured data.This also suggested the existence of significant intersymbolinterferences.
6. Conclusions
We have tested the 220 Mbps QPPM receiver with the newlydeveloped Slik APD and Anadigics ATA12000 preamplifier module.The receiver achieved a BER of 10-6 at an average input opticalsignal power of 4.2 nW (-53.8 dBm), which corresponded to 80incident photons per information bit. The test data are about 1 dBfrom the theoretically predicted performance.
The Slik APD we used had a measured ionization coefficient whichwas four times as large as what was promised [2]. In addition, theoutput of the APD preamplifier had significant ringing at pulsetrailing edges and consequently caused intersymbol interference.The raised cosine filter was approximated with a simple lowpass
filter, which might also cause some degradation in receiverperformance.
7. References
[1] X. Sun and F. M. Davidson, 'Direct detection optical intersatellitelink at 220 Mbps using AIGaAs laser diode and silicon APD with4-ary PPM signaling,1 Interim progress report on NASA grantNAG5-356, 'Optical communication with semiconductor laserdiode' for the period Sept. - Feb. 1990, Dept. ECE, the JohnsHopkins University, Baltimore, MD 21218, March 1990.
[2] A. D. MacGregor, B. Dion, and R. J. Mclntyre 'High sensitivity, highdata rate receivers for ISL using low-noise silicon APD's,' inOptical Space Communication,' proc. SPIE, vol. 1131, ParisFrance, Apr. 24-26, 1989, pp. 176-186.
[3] 'ATA12000 1.2 Gb/s AGC transimpedance amplifier,' Anadigics,Inc., 35 Technology Dr., Warren, NJ 07059 (phone 510-600-5741),Advanced product information, Rev. 1, May 1991.
[4] P. P. Webb, 'Properties of avalanche photodiodes,' RCA Rev., vol.35, pp. 234-278, June 1974.
[5] R. G. Smith and S. D. Personick, 'Receiver design for optical fibercommunication system,' in Semiconductor Devices for OpticalCommunication, Spring-Verlag, New York, 1980, ch. 4.
[6] F. M. Davidson and X. Sun, 'Bandwidth requirements for directdetection optical communication receivers with PPM signaling,1
in Free-Space Laser Communication Technologies III, David L.Begley and Bernard D. Seery, Eds., proc. SPIE 1417, 1991, pp. 75-88.
10
Table 1.Receiver BER vs. Average Input Optical Signal Power
Receiver Average received Average receivedBER Power (nW) photons/bit
3.1e-5 3.09 53.19.6e-6 3.44 64.72.7e-6 3.88 73.01.1e-6 4.15 78.1LOe-6 4.17 78.43.5e-7 4.30 80.98.0e-8 4.87 91.23.3e-8 5.43 1021.0e-8 5.98
Table 2.Parameter Values Used in the Theoretical
Receiver Performance Calculations
binary source data rate 220 MbpsQPPM pulsewidth I=(2X220X1Q6)-1 slaser diode wavelength X=820 nmlaser ON-OFF extinction ratio 300
APD quantum efficiency ri=90%APD surface leakage current ls=15 nAAPD bulk leakage current lb=1.0 pAAPD lonization coefficient keff=0.020
Preamplifier transimpedance Rf=1.5KQequivalent noise temperature Te=700 °K (5.1 pA/VHz)
1 1
CBTJ
X.
Oti\
\
BE
R t
este
rre
ceiv
er
^y k
AOO
u
</) .& 0§2 >o t gN fe 0CM O C
(AQ. i_n a)5^ oo Q- o^j Q. cCVJ O <B
Aco01
T3
3ER
tes
ter
ran
sm
itte
r
s
I
<
0)•o0
T3
i_0)(A(t
COl
E
o
CC
c
•c
a</n
(Daca+-*
0
I
83 'L ,r i
A
7a5 -c) C! T
curr
en
i
i ccc
n.
5kw
T
•>
•
5
)3J5
>«Q.a.a(0
(0
i<-5e* a c0 < 2^. o
0)
1 1j <D . -_.d H— *- Ol
*- 0) iia> * uc — c •— ^
^o*«J *- a CDi= W C T3
3 C 0) COffl o - Sz -o <o i
Q.
0)(/)•«-*(A
3
0)
2oca
o
m
|0>u_
11402 DIGITIZING OSCILLOSCOPEdate: -iO-APR-ri time: iv:J3:50
Cexp:4.l,dig:4.2,dsy:4.OJInstrument ID# B021397
Tek
35mV
Figure 2. Pulse shapes of the laser diode transmittermeasured with a high speed photodiode (invertingoutput).
N
GQin
Q.o^-U)
u0}roE
Q(\J
H2CO
NI
^QQcn2m>
NIZ
2N CDI
cnG UJ
tri-CE<i-tn
. 2 5 ?n= N o
o * >>a E |
^ '«
<D
8 *5 T3 <
§. 2 a,W 3 .C<D O) *-•- CO .
>, ® ^ £
C E '^ ~a) jo w
" 1 2QIII
00
2g>LL
C3 ^3 . ZE = 5
N ^I NZ I
QQ.<
03
O
E
CDQ. CDW 3*- "O
I |O C
Q. ^_
CD 0)CO >^
If
2=3O)
co
114O2 DIGITIZING OSCILLOSCOPEdate: 22-APR-92 time: .13: Iti: U:
Cexp:4.l,digs4.2,dBy«4.0>Instrument ID# B021597
Tfek
57.5mV
flwg(Cl)Fast
M a i n |flvg# >16 | Linear1024 ots
IGHz ! fill Wfms flvg(Cl) | offStatus M a i n
Figure 5. Average output pulse shape from the APDpreamplifier module at 9.45 nW average inputoptical signal power.
biasvoltage
AvantekUTO511/5055-500 MHz,28dB
APD/preampmodule
To timingrecovery ^circuit borad
MiniCircutiZFL1000-LN0.1-1000MHZ23 dB
attenuator-10 dB
outside the receiver case
inside the receiver case
AvantekGPD321/3300.1-1000MHz,21 dB
MiniCircuitZFSC-2-4
ToQPPMdetection ^-circuit board
TrontekP500A-30.1-500 MHz,40 dB.
modifiedAvantekUTC2033
attenuator-6 dB 1
Jo pf
attenuator-9 dB
Figure 6. Modified receiver frontend circuit. Thisshould supersede Fig. 4 of Reference [1].
i!402 DIGITIZING OSCILLOSCOPEdate: -~' -APP -'•? 2 time: - _": -X-: Li 2
:4.lrdig:4.2rdsy:4.0>Instrument ID# B021S97
Tek
312 .5mV
•187. 5mV--36.9ns Sns'd i
582 Rll WfmsStatus
Figure 7. Pulse shapes of input signal to the QPPMreceiver.
11402 DIGITIZING OSCILLOSCOPEsate: ^7-AFP-s; time: 15: 3ci: '.:•:•
Cexp:4.1,dig:4.2,dsy:4.03Instrument ID# B021597
Figure 8. Same as Figure 7 but averaged over 64 traces.
VtAPD output 1
threshold
10G012B
PPM sequenceoutput
DSOU.
load
SR(10G022)
elk
DO,D1=0D2,D3,Cin=l
QO|Q1 JQ4JQ5
load
10G001
PLL#2(16G044)
1
! vcxo: 220MHz -D°l
elk— »
\ rD
Df-f(%10G021)
,0.
8V Vref Vee Vss Vdd Vttii A A A A I
': 90GPPSpower
card#l Vcc to VCXO's
: -8VDC
: ! : : t; +5VDC ; i +28VDC
Figure 9. Modified timing recovery circuit (with anadditional comparator at the input). This shouldsupersede Fig. 6 of References [1].
114O2 DIGITIZING OSCILLOSCOPEdate: ~ ; -:-,.-'P-•?- time: ". .."...;.".
<exp:4.1,diq:4.irdsyj4.C»Instrument ID# B021S97
Ttek
700mV
-1.3V--37.4ns Sns 'div
flvg(C2) I Main |flvg+ >32 I LinearFast fe 1824 nts
Rll Nfms Rvg(C2)Status Main
Figure 10. Average waveform output from the raisedcosine filter (inverting output).
ocLUm
10
10
10
10
10
10
10
10
-3
-4 .
-5 .
-6 .
-7 _
-8 .
-9
-10.
Keff = 0.020Experiment
0 1 2 3 4 5 6 7 8 9 1 0
Received Signal Power (nW)
Figure 11. Receiver BER vs. the average received opticalsignal power. The average APD gain wasoptimized to G=110.